WebGPT: Browser-assisted question-answering with human feedback

Overview

WebGPT is a framework for browser-assisted question-answering that uses human feedback to improve responses. It enables a model to browse the web, gather information, and generate answers while learning from human preferences.

Best for

Best for
Researchers and developers exploring human-in-the-loop QA systems with web access

Use cases

• Building question-answering systems that can search and cite web sources
• Training language models to follow human preferences in information retrieval
• Researching reinforcement learning from human feedback for web-based tasks

Notes

WebGPT is a framework for browser-assisted question-answering that uses human feedback to improve responses. It enables a model to browse the web, gather information, and generate answers while learning from human preferences.

Use cases

• Building question-answering systems that can search and cite web sources
• Training language models to follow human preferences in information retrieval
• Researching reinforcement learning from human feedback for web-based tasks

Pros

• Combines web browsing with human feedback for more grounded answers
• Provides a structured approach to training models on information-seeking tasks
• Open research framework with published methodology

Cons

• Requires human feedback for training, which is costly and time-consuming
• Limited to the capabilities and data available as of 2021
• May be slower than direct QA models due to browser interaction

Indexed from awesome-llm and enriched against its public facts.